System and method to simulate demand and optimize control parameters for a technology platform

Abstract

A system and method are presented for optimizing choices of control parameters. A method includes collecting demand sequences, wherein each demand sequence is associated with a resource managed by a technology platform; processing a demand sequence for a selected resource to calculate an optimized control parameter (CP) value set adapted to manage an automated process within the technology platform, wherein calculating the optimized CP value set for the selected resource includes processing the demand sequence to generate a collection of bootstrapped demand sequences; and processing the bootstrapped demand sequences with a performance prediction process that models the automated process to calculate the optimized CP value set to achieve a best performance metric.

1 . A method for processing demand data for a set of resources in a technology platform, comprising:
 collecting demand sequences, wherein each demand sequence is associated with a resource managed by a technology platform;   processing a demand sequence for a selected resource to calculate an optimized control parameter (CP) value set adapted to manage an automated process within the technology platform, wherein calculating the optimized CP value set for the selected resource includes processing the demand sequence to generate a collection of bootstrapped demand sequences; and   processing the bootstrapped demand sequences with a performance prediction process that models the automated process to calculate the optimized CP value set to achieve a best performance metric;   wherein the bootstrapped demand sequences are generated with a bootstrap process that includes:
 generating a set of Negative Binomial Distribution (NBD) scenarios using an NBD model; 
 determining whether the demand sequence includes banding at a set of demand multiples; and 
 in response to a determination that the demand sequence includes banding, reshaping each of the NBD scenarios to include banding at the set demand multiples and using reshaped NBD scenarios as the bootstrapped demand sequences. 
   
     
     
         2 . The method of  claim 1 , wherein the NBD model is created with a process that includes:
 estimating historical mean demand and forecasting future mean demand from the demand sequence;   aggregating historical demand into blocks of observations and computing demand mean and demand standard deviation in each block;   generating a nonlinear least squares (NLS) regression model that relates demand standard deviation to demand mean;   forecasting future demand standard deviation values using the future mean demand values as input to the NLS regression model; and   using the future mean demand values and future demand standard deviation values as parameters in the NBD model.   
     
     
         3 . The method of  claim 1 , wherein banding is determined with a process that includes:
 compiling an empirical probability mass function (PMF) of the demand sequence;   computing an autocorrelation function of the PMF and identify a nonzero lag with a largest coefficient; and   performing a statistical test of a null hypothesis that a percentage of demand at the nonzero lag and all multiples is the same as the corresponding percentage in a corresponding NBD for all the generated NBD scenarios.   
     
     
         4 . The method of  claim 1 , wherein reshaping each of the NBD scenarios includes:
 for each demand multiple in the demand sequence, determining a demand band that includes a value of the demand multiple and nearby demand values;   determining from the demand sequence an observed proportion of demand for each value in each demand band; and   reassign nearby demand values in each NBD scenario to the value of the demand multiple in a respective demand band according to the observed proportions.   
     
     
         5 . The method of  claim 1 , wherein calculating the optimized CP value set further includes:
 determining whether an initially selected CP value set provides the best performance metric; and   in response to determining that the initially selected CP value set does not provide the best performance metric, evaluating a new neighborhood of CP value set that neighbors and includes the optimized CP value set with the performance prediction process to identify a further optimized CP value set that provides the best performance metric within the new neighborhood.   
     
     
         6 . The method of  claim 1 , wherein the performance prediction process uses a Monte Carlo simulation that models the automated process. 
     
     
         7 . The method of  claim 1 , wherein the performance metrics calculated for all bootstrapped demand sequences for a selected CP value set are averaged to provide a composite performance metric. 
     
     
         8 . The method of  claim 1 , wherein the set of resources are selected from a group consisting of: computing resources, energy resources, web resources, communication resources, physical or virtual components, autonomous vehicles, units of inventory, or Stock Keeping Unit (SKU) identifiers. 
     
     
         9 . The method of  claim 1 , wherein the technology platform is selected from a group consisting of: a cloud computing system, a communication network, a computer network, a control system, a machine, an ERP system, an autonomous vehicle fleet management system, or an inventory management service. 
     
     
         10 . A system, comprising:
 a memory; and   a processor coupled to the memory and configured to process demand data for a set of resources according to a method that includes:
 collecting demand sequences, wherein each demand sequence is associated with a resource managed by a technology platform; 
 processing a demand sequence for a selected resource to calculate an optimized control parameter (CP) value set adapted to manage an automated process within the technology platform, wherein calculating the optimized CP value set for the selected resource includes processing the demand sequence to generate a collection of bootstrapped demand sequences; and 
 processing the bootstrapped demand sequences with a performance prediction process that models the automated process to calculate the optimized CP value set to achieve a best performance metric; 
 wherein the bootstrapped demand sequences are generated with a bootstrap process that includes:
 generating a set of Negative Binomial Distribution (NBD) scenarios using an NBD model; 
 determining whether the demand sequence includes banding at a set of demand multiples; and 
 in response to a determination that the demand sequence includes banding, reshaping each of the NBD scenarios to include banding at the set demand multiples and using reshaped NBD scenarios as the bootstrapped demand sequences. 
 
   
     
     
         11 . The system of  claim 10 , wherein the NBD model is created with a process that includes:
 estimating historical mean demand and forecasting future mean demand from the demand sequence;   aggregating historical demand into blocks of observations and computing demand mean and demand standard deviation in each block;   generating a nonlinear least squares (NLS) regression model that relates demand standard deviation to demand mean;   forecasting future demand standard deviation values using the future mean demand values as input to the NLS regression model; and   using the future mean demand values and future demand standard deviation values as parameters in the NBD model.   
     
     
         12 . The system of  claim 10 , wherein banding is determined with a process that includes:
 compiling an empirical probability mass function (PMF) of the demand sequence;   computing an autocorrelation function of the PMF and identify a nonzero lag with a largest coefficient; and   performing a statistical test of a null hypothesis that a percentage of demand at the nonzero lag and all multiples is the same as the corresponding percentage in a corresponding NBD for all the generated NBD scenarios.   
     
     
         13 . The system of  claim 10 , wherein reshaping each of the NBD scenarios includes:
 for each demand multiple in the demand sequence, determining a demand band that includes a value of the demand multiple and nearby demand values;   determining from the demand sequence an observed proportion of demand for each value in each demand band; and   reassigning nearby demand values in each NBD scenario to the value of the demand multiple in a respective demand band according to the observed proportions.   
     
     
         14 . The system of  claim 10 , wherein calculating the optimized CP value set further includes:
 determining whether an initially selected CP value set provides the best performance metric; and   in response to determining that the initially selected CP value set does not provide the best performance metric, evaluating a new neighborhood of CP value sets that neighbor and include the optimized CP value set with the performance prediction process to identify a further optimized CP value set that provides the best performance metric within the new neighborhood.   
     
     
         15 . The system of  claim 10 , wherein the performance prediction process uses a Monte Carlo simulation that models the automated process. 
     
     
         16 . The system of  claim 10 , wherein the performance metrics calculated for all bootstrapped demand sequences for a selected CP value set are averaged to provide a composite performance metric. 
     
     
         17 . The system of  claim 10 , wherein the CP value set includes a pair a parameters that include at least one of: (Reorder Point, Order Quantity), (Min, Max) or (Review Interval, Order-up-to Level). 
     
     
         18 . The system of  claim 10 , wherein the set of resources are selected from a group consisting of: computing resources, energy resources, web resources, communication resources, physical or virtual components, autonomous vehicles, units of inventory, or Stock Keeping Unit (SKU) identifiers. 
     
     
         19 . The system of  claim 10 , wherein the technology platform is selected from a group consisting of: a cloud computing system, a communication network, a computer network, a control system, a machine, an autonomous vehicle fleet management system, an ERP system, or an inventory management service. 
     
     
         20 . The system of  claim 10 , wherein the performance metric includes a cost.